Search Images Maps Play YouTube News Gmail Drive More »
Sign in
Screen reader users: click this link for accessible mode. Accessible mode has the same essential features but works better with your reader.

Patents

  1. Advanced Patent Search
Publication numberUS7047312 B1
Publication typeGrant
Application numberUS 09/739,309
Publication dateMay 16, 2006
Filing dateDec 18, 2000
Priority dateJul 26, 2000
Fee statusPaid
Publication number09739309, 739309, US 7047312 B1, US 7047312B1, US-B1-7047312, US7047312 B1, US7047312B1
InventorsJames Aweya, Michel Ouellette, Delfin Y. Montuno
Original AssigneeNortel Networks Limited
Export CitationBiBTeX, EndNote, RefMan
External Links: USPTO, USPTO Assignment, Espacenet
TCP rate control with adaptive thresholds
US 7047312 B1
Abstract
The TCP receiver's advertised window (i.e., the receive buffer of a TCP connection) limits the maximum window and consequently the throughput that can be achieved by the sender. Thus, the idea behind TCP rate control is to match the offered network load to the available resources by modifying at an intermediate network device, the receiver's advertised window in TCP acknowledgments returning to the sources. In this disclosure, we propose a new TCP rate control scheme for a shared buffer where the buffer is logically organized into multiple queues. In the scheme, dynamic buffer thresholds are used to ensure efficient and fair usage of buffer memory among the queues. Conventional schemes allocate buffer space to each queue through the use of static buffer thresholds. This can result in unnecessary packet drops which leads to poor network performance since congested or heavily loaded queues cannot gain access to buffers not utilized by lightly loaded queues.
Images(7)
Previous page
Next page
Claims(24)
1. A method for controlling the transmission of data packets through a network by controlling a Transmission Control Protocol (TCP) rate in a network device having a shared buffer with shared buffer space, the method comprising:
organizing a forward data buffer into one or more queues that store at least one forward data packet;
calculating the network device's advertised window size by:
(1) initializing a timer to a predetermined time interval Δt, and an iteration counter to a predetermined initial value n;
(2) sampling a current queue size qi(n) during the predetermined time interval Δt;
(3) calculating a current error signal ei(n) based, at least in part, upon the current queue size qi(n);
(4) calculating the network device's advertised window size Wi(n), based, at least in part, upon the current error signal ei(n) according to the equation: Wi(n)=[Wi(n−1)+αei(n)]W min W max , where α, Wmax, and Wmin, are predetermined parameters;
(5) resetting the timer, upon expiration of the predetermined interval Δt; and
(6) iterating the iteration counter, upon expiration of the predetermined time interval Δt;
providing the network device's advertised window size to a TCP source; and
calculating a dynamic buffer threshold based, at least in part, upon the sum of the queue sizes and the shared buffer space.
2. The method of claim 1 wherein the step of organizing a forward data buffer further comprises:
organizing the forward data buffer into one or more queues with one queue per service class.
3. The method of claim 1 wherein the at least one forward data packet is stored according to its service class.
4. The method of claim 1 wherein the steps of calculating a current error signal ei(n) and calculating the network device's advertised window size further comprise:
filtering the current error signal ei(n) according to the relation: êi(n)=(1−β)êi(n−1)+βei(n), where β is a predetermined parameter; and
calculating the network device's advertised window size Wi(n), based, at least in part, upon the filtered current error signal êi(n) according to the equation: Wi(n)=[Wi(n−1)+αêi(n)]W min W max , where α, Wmax, and Wmin, are predetermined parameters.
5. The method of claim 1 wherein the step of providing the network device's advertised window size to a TCP source further comprises:
carrying information relating to the network device's advertised window size by returning TCP acknowledgements in a receiver's advertised window field.
6. The method of claim 1 wherein the step of providing the network device's advertised window size to a TCP source further comprises:
updating a TCP receiver's advertised window size.
7. The method of claim 6 wherein the step of updating a TCP receiver's advertised window size further comprises:
identifying whether a packet is an ACK packet, and, if not, putting the non-ACK packet in a reverse data buffer;
determining a service class for the identified ACK packet;
reading the TCP receiver's advertised window size (RWrec) and a checksum (RCHKSUM) from the identified ACK packet;
determining whether the TCP receiver's advertised window size RWREC, is less than or equal to the calculated network device's advertised window size Wi(n) and, if not setting a advertised window field in the identified ACK packet equal to the network device's advertised window size Wi(n) and updating the checksum field for the identified ACK packet.
8. The method of claim 1 wherein the step of calculating a dynamic buffer threshold further comprises:
initializing a timer to a predetermined time interval Δs and an iteration counter to a predetermined initial value n;
setting an initial dynamic buffer threshold T(0) equal to a gain constant γ multiplied by a buffer size B and divided by a number of service classes K;
sampling a current queue size qi(n) during the predetermined time interval Δs;
calculating a sum of the sampled current queue size according to the equation: Q ( n ) = i = 1 K q i ( n ) ;
determining whether the sum of the sampled current queue size is less than the product of the gain constant and the buffer size γB;
if so, updating the dynamic buffer threshold according to min{T(n−1)+ΔT, γB}, where ΔT is a step size that controls the rate at which the dynamic buffer threshold changes;
if not, updating the dynamic buffer threshold according to max{T(n−1)−ΔT, Tmin}, where Tmin is a predetermined minimum size for the dynamic buffer threshold;
resetting the timer, upon expiration of the predetermined interval Δs; and
iterating the iteration counter, upon expiration of the predetermined time interval Δs.
9. The method of claim 8 wherein the step of calculating a sum of the sampled current queue size further comprises:
filtering the sum of the sampled current queue size Q(n) according to the relation: {circumflex over (Q)}(n)=(1−φ){circumflex over (Q)}(n−1)+φQ(n), wherein φ is a predetermined parameter.
10. An apparatus for controlling the transmission of data packets through a network by controlling a Transmission Control Protocol (TCP) rate in a network device having a shared buffer with shared buffer space, the apparatus comprising:
a forward data buffer, organized into one or more queues that store at least one forward data packet;
a network device's advertised window size calculation module further comprising:
(1) a timer, initially set to a predetermined time interval Δt, and an iteration counter initially set to a predetermined initial value n;
(2) a current queue size sampler that samples a current queue size qi(n) during the predetermined time interval Δt;
(3) a current error signal calculation module that calculates a current error signal ei(n) based, at least in part, upon the current queue size qi(n); and
(4) a window size calculation module that calculates the network device's advertised window size Wi(n), based, at least in part, upon the current error signal ei(n) according to the equation: Wi(n)=[Wi(n−1)+αei(n)]W min W max , where α, Wmax, and Wmin, are predetermined parameters;
a feed back module that provides the network device's advertised window size to a TCP source; and
a dynamic buffer threshold module that calculates a dynamic buffer threshold based, at least in part, upon the sum of the queue sizes and the shared buffer space.
11. The apparatus of claim 10 wherein the current error signal calculation module further comprises:
a filter module that filters the current error signal ei(n) according to the relation: êi(n)=(1−β)êi(n−1)+βei(n), where β is a predetermined parameter; and
wherein the window size calculation module calculates the network device's advertised window size Wi(n), based, at least in part, upon the filtered current error signal êi(n) according to the equation: Wi(n)=[Wi(n−1)+αêi(n)]W min W max , where α, Wmax, and Wmin, are predetermined parameters.
12. The apparatus of claim 10 wherein the feed back module further comprises:
an advertised window size updating module that updates a TCP receiver's advertised window size.
13. The apparatus of claim 12 wherein the advertised window size updating module further comprises:
an ACK packet identification module that identifies whether a packet is an ACK packet, and, if not, puts the non-ACK packet in a reverse data buffer;
an ACK packet classifier that determines a service class for the identified ACK packet;
an advertised window size reader that reads a TCP receiver's advertised window size (RWrec) and a checksum (RCHKSUM) from the identified ACK packet;
a window size comparison module that determines whether the TCP receiver's advertised window size RWREC, is less than or equal to the calculated network device's advertised window size Wi(n) and, if not sets an advertised window field in the identified ACK packet equal to the calculated network device's advertised window size Wi(n) and updates the checksum field for the identified ACK packet.
14. The apparatus of claim 10 wherein the dynamic buffer threshold module further comprises:
a timer initially set to a predetermined time interval Δs and an iteration counter initially set to a predetermined initial value n;
a current queue size sampler that samples a current queue size qi(n) during the predetermined time interval Δs;
a current queue size calculation module that calculates a sum of the sampled current queue size according to the equation: Q ( n ) = i = 1 K q i ( n ) ,
where K is a number of service classes;
a dynamic buffer threshold determiner that determines whether the sum of the sampled current queue size is less than the product of a gain constant γ and a buffer size B;
and an updating module that updates the dynamic buffer threshold if the sum of the sampled current queue size is less than the product of the gain constant γ and the buffer size B, according to min{T(n−1)+ΔT, γB}, where ΔT is a step size that controls the rate at which the dynamic buffer threshold changes and if the sum of the sampled current queue size is not less than the product of a gain constant γ and a buffer size B, updates the dynamic buffer threshold according to max{T(n−1)−ΔT, Tmin}, where Tmin is a predetermined minimum size for the dynamic buffer threshold.
15. The apparatus of claim 14 wherein the current queue size calculation module further comprises:
a filter that filters the sum of the sampled current queue size Q(n) according to the relation: {circumflex over (Q)}(n)=(1−φ){circumflex over (Q)}(n−1)+φQ(n), wherein φ is a predetermined parameter.
16. An article of manufacture for controlling the transmission of data packets through a network by controlling a Transmission Control Protocol (TCP) rate in a network device having a shared buffer with shared buffer space, the article of manufacture comprising:
at least one processor readable carrier; and
instructions carried on the at least one carrier;
wherein the instructions are configured to be readable from the at least one carrier by at least one processor and thereby cause the at least one processor to operate so as to:
organize a forward data buffer into one or more queues that store at least one forward data packet;
calculate a network device's advertised window size by:
(1) initializing a timer to a predetermined time interval Δt, and an iteration counter to a predetermined initial value n;
(2) sampling a current queue size qi(n) during the predetermined time interval Δt;
(3) calculating a current error signal ei(n) based, at least in part, upon the current queue size qi(n);
(4) calculating the network device's advertised window size Wi(n), based, at least in part, upon the current error signal ei(n) according to the equation: Wi(n)=[Wi(n−1)+αei(n)]W min W max , where α, Wmax, and Wmin, are predetermined parameters;
(5) resetting the timer, upon expiration of the predetermined interval Δt; and
(6) iterating the iteration counter, upon expiration of the predetermined time interval Δt;
provide the network device's advertised window size to a TCP source; and
calculate a dynamic buffer threshold based, at least in part, upon the sum of the queue sizes and the shared buffer space.
17. The article of manufacture of claim 16 wherein the instructions are configured to be readable from the at least one carrier by at least one processor and thereby cause the at least one processor to operate so as to:
organize the forward data buffer into one or more queues with one queue per service class.
18. The article of manufacture of claim 16 wherein the instructions are configured to be readable from the at least one carrier by at least one processor and thereby cause the at least one processor to operate so as to:
store the at least one forward data packet according to its service class.
19. The article of manufacture of claim 16 wherein the instructions are configured to be readable from the at least one carrier by at least one processor and thereby cause the at least one processor to operate so as to:
filter the current error signal ei(n) according to the relation: êi(n)=(1−β)êi(n−1)+βei(n), where β is a predetermined parameter; and
calculate the network device's advertised window size Wi(n), based, at least in part, upon the filtered current error signal êi(n) according to the equation: Wi(n)=[Wi(n−1)+αêi(n)]W min W max , where α, Wmax, and Wmin, are predetermined parameters.
20. The article of manufacture of claim 16 wherein the instructions are configured to be readable from the at least one carrier by at least one processor and thereby cause the at least one processor to operate so as to:
carry information relating to the network device's advertised window size by returning TCP acknowledgements in a receiver's advertised window field.
21. The article of manufacture of claim 16 wherein the instructions are configured to be readable from the at least one carrier by at least one processor and thereby cause the at least one processor to operate so as to:
update a TCP receiver's advertised window size.
22. The article of manufacture of claim 21 wherein the instructions are configured to be readable from the at least one carrier by at least one processor and thereby cause the at least one processor to operate so as to:
identify whether a packet is an ACK packet, and, if not, put the non-ACK packet in a reverse data buffer;
determine a service class for the identified ACK packet;
read a TCP receiver's advertised window size (RWrec) and a checksum (RCHKSUM) from the identified ACK packet;
determine whether the TCP receiver's advertised window size RWREC, is less than or equal to the calculated network device's advertised window size Wi(n) and, if not setting an advertised window field in the identified ACK packet equal to the calculated network device's advertised window size Wi(n) and updating the checksum field for the identified ACK packet.
23. The article of manufacture of claim 16 wherein the instructions are configured to be readable from the at least one carrier by at least one processor and thereby cause the at least one processor to operate so as to:
initialize a timer to a predetermined time interval Δs and an iteration counter to a predetermined initial value n;
set an initial dynamic buffer threshold T(0) equal to a gain constant γ multiplied by a buffer size B and divided by a number of service classes K;
sample a current queue size qi(n) during the predetermined time interval Δs;
calculate a sum of the sampled current queue size according to the equation: Q ( n ) = i = 1 K q i ( n ) ;
determine whether the sum of the sampled current queue size is less than the product of the gain constant and the buffer size γB;
if so, updating the dynamic buffer threshold according to min{T(n−1)+ΔT, γB}, where ΔT is a step size that controls the rate at which the dynamic buffer threshold changes;
if not, updating the dynamic buffer threshold according to max{T(n−1)−ΔT, Tmin}, where Tmin is a predetermined minimum size for the dynamic buffer threshold;
reset the timer, upon expiration of the predetermined interval Δs; and
iterate the iteration counter, upon expiration of the predetermined time interval Δs.
24. The article of manufacture of claim 23 wherein the instructions are configured to be readable from the at least one carrier by at least one processor and thereby cause the at least one processor to operate so as to:
filter the sum of the sampled current queue size Q(n) according to the relation: {circumflex over (Q)}(n)=(1−φ){circumflex over (Q)}(n−1)+φQ(n), wherein φ is a predetermined parameter.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS

This patent application claims priority based on U.S. Provisional Patent Application No. 60/221,338, filed Jul. 26, 2000, titled, “TCP RATE CONTROL WITH ADAPTIVE THRESHOLDS”, which is hereby incorporated by reference herein in its entirety.

FIELD OF THE INVENTION

The present invention relates generally to data packet flow in a network and, more particularly, to a technique for controlling a transmission control protocol (TCP) rate in a network having a shared buffer.

BACKGROUND OF THE INVENTION

The primary protocol for congestion control in the Internet is TCP. The rapid growth of applications and services over the Internet has made the performance of TCP a major issue in data networking. TCP performs its own data recovery and flow control, and makes no assumption on how the network processes the data it sends. TCP congestion control as well as error recovery are implemented by a sliding window. The TCP flow control mechanism is meant to slow down the source rate when the network becomes congested, but currently, TCP has no direct way of knowing when the network is congested. It can only indirectly detect congestion by keeping track of how many packets are lost. When packets do get lost, the loss indicates that some queue in the network might have overflowed. Every time TCP detects a packet loss, it reduces its source rate to alleviate the congestion that could have caused the packet loss.

In a high-latency network environment, the window flow control mechanism of TCP may not be very effective because it relies on packet loss to signal congestion, instead of avoiding congestion and buffer overflow. The basic problem is that TCP does not communicate directly with the network elements to be given its optimal or assigned traffic rate. By the time the TCP source starts decreasing its rate because of packet loss, the network has already been overly congested. This problem exists because the design of TCP currently only considers the flow control needs of the receiver. It does not consider the flow control needs of intermediate hops in the network. Overflow in the network itself would be detected by the sender through timeouts or through acknowledgment arrival patterns. This model has problems coping with shared multi-hop networks, where the cause of packet loss is within intermediate hops in the network.

Thus, it is generally accepted that there is a limit to how much control can be accomplished by the congestion control mechanisms in the end systems. Some mechanisms are needed in the routers to complement the endpoint congestion avoidance mechanisms. As a result, network layer enhancements such as scheduling mechanisms (e.g., Fair Queuing) and packet drop policies (e.g., Random Early Detection (RED)) have been proposed which are aimed at improving fairness and throughput.

Ideally, the source should respond to congestion (in the network and at the destination) before it occurs rather than acting when it is too late. Thus, one method to make TCP sensitive to network congestion and buffer overflows is to allow network elements between a source and destination to modify the receiver's advertised window in TCP acknowledgments returning to the source. Through this the network becomes an active (and not a passive) participant in controlling congestion and buffer overflows, similar in spirit to the intelligent congestion control techniques such as RED or RED combined with Explicit Congestion Notification (ECN). Thus, the motivation for TCP rate control and similar schemes, is the need for network control policies which do not necessarily resort to packet discard as the only means for congestion control.

Most of the existing window adjustment schemes are based on heuristics and their performance under a wide range of networking conditions are not clearly understood. The drawback of one scheme (see e.g., R. Satyovolu, K. Duvedi, and S. Kaylanaraman, “Explicit Rate Control of TCP Applications”, ATM Forum Document, ATM-Forum/98-0152R1, February 1998.) is that the optimal setting of the window size to be fed back to the TCP sources, requires accurate knowledge of the round-trip time (RTT) estimates. Also, such information is usually not available at network elements. Another scheme (see e.g., P. Narvaez and K-Y Siu, “An Acknowledgement Bucket Scheme for Regulating TCP Flow Over ATM”, Proc. Globecom '97, Phoenix, Ariz., 1997.) entails too much per-flow state at access nodes, essentially requiring per-flow window tracking.

In view of the foregoing, it would be desirable to provide a technique for data packet flow in a network which overcomes the above-described inadequacies and shortcomings. More particularly, it would be desirable to provide a technique for controlling TCP rate in a network in an efficient and cost effective manner.

SUMMARY OF THE INVENTION

According to the present invention, a technique for controlling the transmission of data packets through a network by controlling a TCP rate in a network device having a shared buffer with shared buffer space, is provided. In one embodiment, the technique is realized by first organizing a forward data buffer in a network device into one or more queues that store at least one forward data packet. A network device's advertised window size is then calculated by implementing an integral control algorithm that uses information pertaining to the one or more queues. The network device's advertised window size is then fed back to a TCP source. A dynamic buffer threshold is also implemented to ensure efficient and fair use of the forward data buffer memory.

In accordance with other aspects of the present invention, the forward data buffer may be organized into one or more queues with one queue per service class. There can be any integer number K of such service class queues. In addition, the forward data packet may be stored according to its service class.

In accordance with still other aspects of the present invention, the network device's advertised window size may be calculated by first initializing a timer to a predetermined time interval At. A current queue size q,(n) for queue i,i=1,2, . . . K,, is then sampled during the predetermined time interval nΔt, where (n=0, 1, 2, . . . ) is a current time interval index or iteration counter and represents the number of time intervals Δt since the beginning of the process of network device's advertised window calculation. A current error signal ei(n) based, at least in part, upon the current queue size qi(n), is then calculated. In some embodiments the current error signal ei(n) may be filtered according to the relation: êi(n)=(1−β)êi(n−1)+βei(n), where β(0<β<1) is a predetermined parameter and êi(n) is the filtered current computed error signal. Next, a network device's advertised window size Wi(n) for queue i may be calculated, based, at least in part, upon the current error signal ei(n) according to the equation: Wi(n)=[Wi(n−1)+αei(n)]W min W max , where α, Wmax and Wmin, are predetermined parameters. In some embodiments, this calculation of the network device's advertised window size Wi(n) may be calculated using the filtered current error signal êi(n) in the equation Wi(n)=[Wi(n−1)+αêi(n)]W min W max , where α, Wmax, and Wmin, are predetermined parameters. Finally, the timer may be reset for the calculation to resume anew.

In accordance with still other aspect of the present invention the feedback to the TCP source may be accomplished by first updating a TCP receiver's advertised window size. In some embodiments updating a TCP receiver's advertised window size is accomplished by first identifying whether a packet is an acknowledgment (ACK) packet, and, if not, putting the non-ACK packet in a reverse data buffer. A service class i(i=1,2, . . . k) for the identified ACK packet is then determined. An advertised window size (RWrec) and a checksum (RCHKSUM) are then read from the identified ACK packet. Then, a determination of whether the advertised window size RWREC, is less than or equal to the calculated network device's advertised window size Wi(n) is carried out. If the advertised window size RWREC, is not less than or equal to the calculated network device's advertised window size Wi(n), then an advertised window field in the identified ACK packet is set equal to the calculated network device's advertised window size Wi(n) and the checksum field for the identified ACK packet is updated.

In accordance with still other aspect of the present invention a dynamic buffer threshold is implemented by first calculating a dynamic buffer threshold for the forward data buffer. In some embodiments, calculating a dynamic buffer threshold is accomplished by first initializing a timer to a predetermined time interval Δs. Then an initial dynamic buffer threshold T(n=0) is set equal to a parameter γε(0,1] multiplied by a buffer size B and divided by a number of service classes K. A current queue size qi(n) is sampled during the predetermined time interval nΔs. Then a sum of the sampled current queue size is calculated according to the equation:

Q ( n ) = i = 1 K q i ( n ) .
In some embodiments, the sum of the sampled current queue size Q(n) is filtered according to the relation: Q(n)=(1−φ){circumflex over (Q)}(n−1)+φQ(n), wherein φ(0<φ<1) is a predetermined parameter and {circumflex over (Q)}(n) is the filtered sum of queue sizes. Then a determination of whether the sum of the sampled current queue size is less than the shared buffer size γB is made. If it is, the dynamic buffer threshold is updated according to T(n)=min{T(n−1)+ΔT,γB}, where ΔT is a step size that controls the rate at which the dynamic buffer threshold changes. If it is not, the dynamic buffer threshold is updated according to T(n)=max{T(n−1)−ΔT,Tmin}, where Tmin is a predetermined minimum size for the dynamic buffer threshold. Finally, the timer may be reset for the calculation to resume anew.

The present invention will now be described in more detail with reference to exemplary embodiments thereof as shown in the appended drawings. While the present invention is described below with reference to preferred embodiments, it should be understood that the present invention is not limited thereto. Those of ordinary skill in the art having access to the teachings herein will recognize additional implementations, modifications, and embodiments, as well as other fields of use, which are within the scope of the present invention as disclosed and claimed herein, and with respect to which the present invention could be of significant utility.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to facilitate a fuller understanding of the present invention, reference is now made to the appended drawings. These drawings should not be construed as limiting the present invention, but are intended to be exemplary only.

FIG. 1 is a TCP rate control scheme according to one embodiment of the present invention.

FIG. 2 is a block diagram of a feedback system according to one embodiment of the invention.

FIG. 3 is a block diagram of a feedback system providing anti-windup in a controller according to one embodiment of the invention.

FIG. 4 is a flow diagram for a method of computing a network device's advertised window size for the connections in each service class according to one embodiment of the invention.

FIG. 5 is a flow diagram for a method of updating a TCP receiver's advertised window field according to one embodiment of the invention.

FIG. 6 is a flow diagram for a method of computing a dynamic threshold according to one embodiment of the invention.

FIG. 7 is a schematic diagram of an overall TCP rate control scheme according to one embodiment of the invention.

DETAILED DESCRIPTION OF EXEMPLARY EMBODIMENT(S)

One embodiment of the present invention enables a simple TCP rate control scheme for a shared-buffer where the buffer is organized into multiple queues (e.g., one queue per service class). In the shared-buffer, packets can be stored according to service class as a linked-list and served using any service discipline (e.g., round robin, weighted round robin, weighted fair queuing, etc.). Some embodiments of the TCP rate control scheme use an integral control algorithm to compute the optimal window size to be fed back to the TCP sources. In some embodiments, dynamic buffer thresholds are used to ensure efficient and fair usage of buffer memory among the queues. Conventional schemes allocate buffer space to each queue through the use of static buffer thresholds which can result in unnecessary packet drops and lead to poor network performance since congested or heavily loaded queues cannot gain access to buffers not utilized by lightly loaded queues.

A shared-buffer without good buffer management procedures may not perform well under overload conditions, especially in regard to fairness. The statistical sharing of the buffer memory introduces the possibility of a small group of queues hogging most of the memory, preventing packets destined for less utilized queues from gaining their fair shares of memory or even blocked completely. Conventional solutions to such problems include setting some restrictions on the amount of buffering a queue can use (i.e., each queue has a threshold for the maximum number of buffers it can access, each queue is guaranteed a minimum number of buffers, etc.).

In some embodiments of the dynamic threshold scheme described here, buffer thresholds are dynamically adjusted according to the changing traffic load. Through this congested queues are allowed to gain access to buffers not utilized by lightly loaded queues as long as their queues do not exceed their fair share of the memory. This helps to improve the overall buffer utilization while ensuring fair sharing of buffer memory among the different queues.

The TCP rate control scheme in the present invention defines the control error as the difference between a dynamic threshold or reference queue size and the actual queue size. The present invention utilizes dynamic thresholds for each queue and the window control algorithm used to compute the network advertised window size has a control gain which determines the stability and response of the system and how well the control algorithm operates in steady-state. The sampling and control intervals in the present invention are fixed to enable efficient and simple processor implementation in discrete time steps. In addition, the present invention involves filtering of the control errors (or queue sizes) which has the important benefit of allowing traffic bursts to be queued without being unnecessarily discarded. This is because congestion is detected by comparing the average queue size to a pre-determined threshold. In effect, only the average queue size is controlled, thus allowing to some extent, transient traffic bursts to be accommodated in the queue.

A functional block diagram illustrating the proposed TCP rate control scheme is given FIG. 1. In this embodiment, the forward data buffer 10 is organized into multiple queues. For example, buffer 10 may comprise queues 12, 14 and 16, wherein the buffer space is shared among a number of service classes (e.g., class 1, class 2, . . . , class K). For example, in a network device with a shared buffer, packets destined to a particular service class can be classified (e.g., by packet classifier 18) and stored according to service class as a linked-list in the shared buffer space.

In the reverse direction, ACK packet identification is implemented via ACK packet identifier 20, so that ACK packets can be separated from normal (non-ACK) data traffic. To identify an ACK packet, the network device checks the ACK bit in the TCP packet header to determine if the ACK bit is set. An ACK packet has its ACK bit set by the TCP receiver. ACK information might also be contained (piggybacked) in a data packet transmitted from a TCP receiver to a sender. ACK packet classifier 22 filters and separates ACK packets according to their individual classes (e.g., class 1, class 2, . . . , class K) so that their receivers' advertised windows and checksums can be updated, (e.g., at window and checksums update module 24). Also, in the reverse direction, ACK first scheduling can be used to give priority to ACK packets over reverse data packets since the number of ACK packets is normally small compared to data packets.

A wide range of traffic classes can be supported depending on the implementation complexity allowed. The criteria for classifying or filtering the ACK traffic flows can be based on source/destination Internet Protocol (IP) addresses, TCP port numbers, the Type of Service (TOS) field (or DS field), and any higher level IP packet attributes.

The architecture illustrated in the embodiment shown FIG. 1 implements the following computational functions:

  • (1) A window and threshold computations module 26 for computing the optimal window size per connection in each service class queue at the network device (i.e., the network device's advertised window size) and the buffer thresholds for the forward data buffer.
  • (2) A window and checksum update module 24 for updating the TCP receiver's advertised window size and checksum in each returning ACK packet in each service class.
    Each module is described in more detail below.

The window control method of some embodiments is a mechanism for congestion control (e.g., controlling the queue size) at a line interface of a network device. The method controls congestion by modifying at an intermediate network element, the receiver's advertised window in TCP acknowledgments returning to the sources. A critical parameter in the window control scheme is determining the optimal network device's advertised window size per connection. The network device's window size represents the maximum number of packets that a TCP connection can have in transit (i.e., unacknowledged by a receiver) at a time. In determining the network device's window size per connection, one embodiment of the TCP rate control scheme tries to match the sum of the windows of all the active TCP connections sharing the network node to the network bandwidth-delay product plus the available buffering space, thus avoiding packet losses whenever possible. The method takes its decision of modifying the receiver's advertised window of an incoming ACK packet so that the queue occupancy level is kept at a given target level, thereby eliminating buffer underflow and overflow as much as possible.

For the multiple TCP connections that share a common service class queue, the window control method attempts to match the aggregate window sizes of all active TCP flows to the bandwidth-delay product plus the available buffering space at the queue while at the same time providing all the connections in the queue with similar feedback to achieve fairness in window size allocation. The feedback may be carried by returning TCP acknowledgments in the receiver's advertised window field. If the current value in the receiver's advertised window field, which is set by the destination system, exceeds the feedback value computed in the network device, the receiver's advertised window may be marked down to the feedback value. The computed feedback value may provide an upper bound for the TCP congestion window maintained at the source in order to limit packet losses in the network device.

In some embodiments, the forward data buffer 10 (e.g., a shared memory) has a capacity of B packets shared by the K service classes. The queue (e.g. 12, 14, 16) size in the network device (e.g., switch or router) is assumed to be sampled every Δt units of time (e.g., seconds), and the window control method provides every Δt units of time, a new value of the network device's advertised window size per connection Wi for service class i (i=1,2, . . . ,K). Therefore, Δt is the sampling/control interval of the system. The control system according to some embodiments of the invention may be described in discrete units of time as follows.

Let qi(n) denote the actual queue size of class i, i=1,2, . . . ,K, at discrete time n, where n=1Δt, 2Δt, 3Δt, . . . , and let T denote the target buffer occupancy per service class at the network device. Next, determine a window size Wi which will drive the queue size of class i, to this target buffer occupancy. Then, adapt Wi to react to the actual queue dynamics experienced at the network device using a simple control mechanism. For example, if the actual queue size qi is smaller than the target queue size Ti Wi is increased to make aggressive usage of the available resources, and vice versa if the actual queue size is high. The goal of the controller is therefore to adapt Wi so that the magnitude of the error signal
e i(n)=T−q i(n), i=1,2, . . . ,K,
is kept as small as possible.

In some embodiments a control theoretic approach may be implemented as is explained using the simple feedback control loop shown in FIG. 2. Generally, such a system in the TCP rate control context will consist of: a control target or reference level 200 (this corresponds to a target buffer occupancy at the network device); an output or controlled variable 260 (this corresponds to the actual queue size at the network device); a process or plant 220 (this is the process or device(s) which produces the output and corresponds to all the TCP data sources and the network devices that receive and process the data); a controller or decision-making device 230 (this is where adjustments to be made in the level of process or plant operation are calculated or are decided upon and corresponds to the module which computes the TCP window size to be communicated to the sources); a feedback device 240 (this is the device which measures some feature of the output and processes the information in preparation for comparison with the control target); and a comparator or detector 250 (this is the device which actually compares the recorded output with the control target).

The purpose of the control system is to keep the process output 260 close to the control target 200 in spite of disturbances. This is achieved by the feedback 240 loop, which works as follows. Assume that the system is in equilibrium and that a disturbance occurs so that the process output 260 becomes larger than the control target 200. The error 270 is then negative and the control variable 210 decreases which in turn causes the process output 260 to decrease. This type of feedback 240 is called negative feedback, because the control variable 210 (or manipulated variable) moves in direction opposite to the process output 260.

The controller 230 has several parameters that can be adjusted. The control loop performs well if the parameters are chosen properly. It performs poorly otherwise, e.g., the system may become unstable.

Some embodiments of the invention implement integral control. The main function of an integral or summation control is to make sure that the process output 260 agrees with the target 200 in steady-state. With integral action, a small positive error 270 (e) will always lead to an increasing control signal, and a negative error 270 will give a decreasing control signal no matter how small the error is. Integral control gives a control action that is proportional to the time integral of the error 270. This ensures that the steady-state error 270 becomes zero. The integral action is given by

I ( t ) = γ T 1 0 t e ( t ) τ ,
where γ is a gain constant and Ti is the integral time.
It follows that

I t = γ T 1 e .
Approximating the derivative by the backward difference gives

I ( k Δ t ) - I ( ( k - 1 ) Δ t ) Δ t = γ T 1 e ( k Δ t ) , k = 0 , 1 , 2 , ,
where Δt is the sampling period. This leads to the following recursive equation for the integral action

I ( k Δ t ) - I ( ( k - 1 ) Δ t ) + γ Δ t T 1 e ( k Δ t ) = I ( ( k - 1 ) Δ t ) + α e ( k Δ t )
or
I(n)=I(n−1)+α e(n)
where n=0,1,2, . . . , is a discrete time index.

Due to the burstiness of the network traffic and other perturbations, this error signal may be highly fluctuating, so that some low pass filtering may be desirable. Thus, the first part of the controller 230 may be a discrete-time first-order low-pass filter with gain 0<β<1. The filtered error signal is given by
ê i(n)=(1−β)ê i(n−1)+βe i(n), i=1,2, . . . , K.

In some embodiments, the second part of the controller is then the incremental adaptation of the network device's advertised window Wi proportional to the filtered error signal
W i(n)=W i(n−1)+αê i(n), i=1,2, . . . , K,
where α is a control gain.

For network traffic control with a wide range of operating conditions, it may happen that there is a persistent control error 270 (or mismatch) between the control target 200 (T) and the observed system state (q) 260. This persistent error 270 may be due to rate constraints at certain network nodes. For example, the errors 270 may be due to rate constrained traffic source(s), rate constraints due to downstream bottlenecks, or rate reduction due to upstream and/or downstream packet losses. If a controller 230 with integral action is used, the error 270 will continue to be integrated. This means the integral term (and as a result the control variable 210) may become very large or it “winds up”. It is then required that the error 270 has opposite sign for a long period before things return to normal.

One way to avoid integrator windup is to introduce limiters or integrator anti-windup on the variables that can windup. The effect of anti-windup is to reduce both the overshoot and the control effort in the feedback control system as represented in FIG. 2. Implementatation of such anti-windup schemes is preferable in any practical implementation of integral control. Omission of this technique may lead to deterioration of response and even stability. FIG. 3 shows how a saturation 300 can limit the control variable directly providing anti-windup. However, too hard a limitation will cause an unnecessary limitation of the control action, and too weak a limitation will cause windup. So the limits must be selected carefully. The upper bound Wmax can be set to a maximum end system advertised window such as the 16-bit window size TCP protocol field (65536 bytes) or any suitable value.

Thus, the control equation becomes
W i(n)=[W i(n−1)+αê i(n)]W min W max
where the minimum window size is Wmin≧0.

Flowcharts illustrating embodiments of the window computations and the TCP advertised window update at a network node are presented in FIG. 4, and FIG. 5, respectively. As shown in FIG. 4, one embodiment of the invention may compute a network device's advertised window size in the following manner. At step 400 initial parameters may be set. For example, the time interval for taking measurements Δt, the filtered error signal ê,(n), and the advertised window size Wi(n), may be set to initial values (e.g., êi(n)=0, n=0, etc.). At step 410, upon expiration of an interval Δt, the timer may be reset to Δt, and n may be iterated to n=n+1. At step 420 a current queue size qi(n) may be sampled. At step 430 a current error may be computed (e.g., ei(n)=T(n)−qi(n)). In some embodiments, an optional step 440 may be implemented to compute a filtered error (e.g., êi(n)(1−β)êi(n−1)+βei(n)). In some embodiments, if a filtered error is not calculated, then the current computed error êi(n)=ei(n) is used. At step 450, the current window sizes may be computed (e.g., Wi(n)=[Wi(n−1)+αêi(n)]W min W max ). In some embodiments, the process may loop back to step 410 for another iteration.

FIG. 5 shows TCP receiver's advertised window field update according to one embodiment of the invention. The process may initiate at step 500 with the arrival of a data packet at an ACK packet identification unit (e.g., ACK packet identifier 20). At step 510 the process determines whether the data packet is an ACK packet. If it is determined that the packet is not a ACK packet, at step 520 the packet is placed in a reverse data buffer (e.g., reverse data buffer 28) and ultimately forwarded to a multiplexer (e.g., Mux 30) at step 580. If the packet is determined to be an ACK packet, a determination of the ACK packet class is made at step 530 (e.g., by packet classifier 22). At step 540 the TCP receiver's advertised window (e.g., RWREC) and checksum (RCHKSUM) are read from the ACK packet (e.g., by window and checksum update module 24). Step 550 is to determine the read TCP receiver's advertised window size against the network device's advertised window size (e.g., RWREC≦Wi(n)?). If the read TCP receiver's advertised window RWREC is less than the network device's advertised window size Wi(n), then the ACK packet is forwarded to a multiplexer (e.g., Mux 30) at step 580. If the read TCP receiver's advertised window RWREC is greater than the network device's advertised window size Wi(n), then the ACK packet's TCP receiver's advertised window size is set to equal Wi(n) at step 560. At step 570 the checksum field of the ACK packet is updated (e.g., by window and checksum update module 24) using RCHKSUM, RWREC, and Wi(n). The updated ACK packet is forwarded to a multiplexer (e.g., Mux 30) at step 580.

In some embodiments of TCP rate control, the TCP implementations in the end systems do not need to be modified, nor is it required to modifying the TCP protocol itself. Updating the window size in returning ACKs is done in a manner transparent to TCP. The processing performed by the network device on each TCP acknowledgment consists of updating the TCP receiver's advertised window and the checksum fields. The checksum can be updated from knowing only the previous and the new values of the advertised window field, as well as the old checksum value. For example, an efficient technique for incrementally updating the checksum is given in A. Rijsinghani, “Computation of the Internal Checksum via Incremental Update,” IETF RFC 1624, May 1994.

Listed below are the parameters that may affect the control performance.

  • (1) Sampling intervals (Δt and Δs): Δt and Δs are the time intervals for taking measurements and applying the computed Wi, i=1,2, . . . , K and T controls, respectively, (e.g., Δt=10 packet transmission time and Δs=100 msec).
  • (2) Control gain (α): This controls the reaction speed and stability of the control system.
  • (3) Filter gain (β): This controls the reaction speed of the filter. The recommended value for some embodiments is β=0.002.
  • (4) Control target (i.e., computed common buffer threshold, T). This sets the average buffer utilization level and average queuing delay for each service class, since the average queue size evolves around this value.
  • (5) Buffer size (B). This is the buffer size allocated to the connections or flows anticipated in the system and is an issue related to buffer dimensioning and traffic engineering. B influences packet losses and system utilization level and is typically selected to be at least equal to a bandwidth-delay product worth of data.

Let qi(n) denote the queue size of class i at discrete time n, i=1,2, . . . , K. The queue size q,(n) may fluctuate as the incoming traffic load and/or the service bandwidth allocated at class i vary with time. To ensure fair sharing of the buffer space B, some embodiments adopt the policy that each class should be able to access at least B/K buffers (i.e., the fair share) when it is overloaded. On the other hand, to efficiently utilize the buffer, (i.e., to achieve maximum buffer sharing) an overloaded class must be able to use any available buffer memory when such buffers are not used by other classes. Thus, we use a common threshold T(n) for all K service classes. One embodiment of a method for computing dynamic thresholds is shown in FIG. 6 where, at discrete time instances, the threshold computations are as follows:

  • (1) Initialization:

T ( 0 ) γ B K

  • (2) After every interval of Δs time units (in seconds), compute:

T ( n ) { min { T ( n - 1 ) + Δ T , γ B } , if i = 1 K q i ( n ) < γ B max { T ( n - 1 ) - Δ T , T min } , if i = 1 K q 1 ( n ) γ B .

The parameter Tmin≧0 represents the minimum buffer threshold statically set for each service class. This is a user-configurable per-class quota to allow the network manager to assign the minimum number of buffers that each class can consume, thus preventing starvation of buffer resources. Through this, the dynamic buffering scheme allows the system to adapt intelligently to various traffic conditions and gives the network manager the ability to fine tune the network to different traffic characteristics. The step size ΔT>0 controls the rate at which the buffer threshold changes. The buffer space (1−γ)B is the slack buffer capacity used to absorb incoming traffic bursts. Thus, the parameter γ ε(0,1] controls the extent to which the buffer space B is completely or partially shared: γ=1 results in complete sharing and 0<γ<1 in partial sharing. If all classes are active, the slack buffer capacity per class is (1−γ)B/K.

In steady-state, the common threshold T is given by T=γB/Nα, where 1≦Nα≦K is the effective number of active classes in the system. In other words, the effective number of active classes at a given time instant is given by Nα=γB/T.

One embodiment of the computation of the dynamic buffer thresholds is illustrated in FIG. 6.

A high-level description of the overall TCP rate control scheme taking into consideration all of the above discussions is given in FIG. 7.

As shown in FIG. 6, one embodiment of a method for dynamic threshold computation may initiate at step 600 by setting initial parameter values

( e . g . , set timer to Δ s , initialize T ( 0 ) = γ B K ) .
At step 610, upon expiration of an interval Δs, the timer may be reset to Δs and n may be iterated to n=n+1. At step 620, the current queue size qi(n) (e.g., queues 12, 14, 16) may be sampled. At step 630 a sum of sampled queue sizes may be calculated

( e . g . , Q ( n ) = i = 1 K q 1 ( n ) ) .
Some embodiments of the invention may optionally filter the sum of queue sizes to reduce fluctuations as described above (e.g., {circumflex over (Q)}(n)=(1−φ){circumflex over (Q)}(n−1)+φQ(n)). If the sum of queue sizes is not filtered then {circumflex over (Q)}(n) is set equal to Q(n). At step 640 a comparison of the sum of queue sizes to the size of the shared buffer space is made (e.g., {circumflex over (Q)}(n)<γ B). If the queue size sum {circumflex over (Q)}(n) is less than the shared buffer space γ B, then, at step 650, the queue threshold is updated to a minimum value (e.g., T(n)=min{T(n−1)+ΔT,γ B}). If the queue size sum {circumflex over (Q)}(n) is greater than the shared buffer space γ B, then the queue threshold is updated to a maximum value (e.g., T(n)=max{T(n−1)−ΔT,Tmin}). In some embodiments the process may return to step 610 for another iteration.

Effective control of TCP connections sharing a common bottleneck link, can be achieved by modifying at the bottleneck link, the receiver's advertised window in TCP acknowledgments returning to the senders. This is because the dynamics of TCP, specifically, a mismatch between the TCP window and the bandwidth-delay product of the network can cause the network switches or routers to accumulate large queues, resulting in buffer overflows, reduced throughput, unfairness and underutilization. As described above a TCP rate control scheme may be used for a shared-buffer with multiple queues where dynamic thresholds are used to ensure efficient and fair usage of buffer memory. The window control algorithm modifies the receiver's advertised window of an incoming ACK packet so that the queue occupancy level is kept at a given target level, thereby eliminating buffer underflow and overflow as much as possible.

The present invention is not to be limited in scope by the specific embodiments described herein. Indeed, various modifications of the present invention, in addition to those described herein, will be apparent to those of ordinary skill in the art from the foregoing description and accompanying drawings. Thus, such modifications are intended to fall within the scope of the following appended claims. Further, although the present invention has been described herein in the context of a particular implementation in a particular environment for a particular purpose, those of ordinary skill in the art will recognize that its usefulness is not limited thereto and that the present invention can be beneficially implemented in any number of environments for any number of purposes. Accordingly, the claims set forth below should be construed in view of the full breath and spirit of the present invention as disclosed herein.

Patent Citations
Cited PatentFiling datePublication dateApplicantTitle
US5764235 *Mar 25, 1996Jun 9, 1998Insight Development CorporationComputer implemented method and system for transmitting graphical images from server to client at user selectable resolution
US5822537 *Apr 5, 1996Oct 13, 1998At&T Corp.Multimedia networked system detecting congestion by monitoring buffers' threshold and compensating by reducing video transmittal rate then reducing audio playback rate
US5917822 *Apr 12, 1996Jun 29, 1999Xerox CorporationMethod for providing integrated packet services over a shared-media network
US6005599 *Dec 23, 1994Dec 21, 1999Hitachi, Ltd.Video storage and delivery apparatus and system
US6028842 *Dec 23, 1996Feb 22, 2000Nortel Networks CorporationDynamic traffic conditioning
US6038216 *Nov 1, 1996Mar 14, 2000Packeteer, Inc.Method for explicit data rate control in a packet communication environment without data rate supervision
US6084919 *Jan 30, 1998Jul 4, 2000Motorola, Inc.Communication unit having spectral adaptability
US6091947 *Mar 13, 1998Jul 18, 2000Sumner; Terence EdwardMethod and apparatus for accepting and conveying a voice mail message to a mobile unit in a wireless telephone system
US6185221 *Nov 9, 1998Feb 6, 2001Cabletron Systems, Inc.Method and apparatus for fair and efficient scheduling of variable-size data packets in an input-buffered multipoint switch
US6185625 *Dec 20, 1996Feb 6, 2001Intel CorporationScaling proxy server sending to the client a graphical user interface for establishing object encoding preferences after receiving the client's request for the object
US6215769 *Oct 7, 1998Apr 10, 2001Nokia Telecommunications, Inc.Enhanced acknowledgment pacing device and method for TCP connections
US6219713 *Jul 7, 1998Apr 17, 2001Nokia Telecommunications, OyMethod and apparatus for adjustment of TCP sliding window with information about network conditions
US6240463 *Nov 24, 1998May 29, 2001Lucent Technologies Inc.Router placement methods and apparatus for designing IP networks with performance guarantees
US6252851 *Mar 27, 1998Jun 26, 2001Massachusetts Institute Of TechnologyMethod for regulating TCP flow over heterogeneous networks
US6269078 *Apr 4, 1997Jul 31, 2001T. V. LakshmanMethod and apparatus for supporting compressed video with explicit rate congestion control
US6301258 *Dec 4, 1997Oct 9, 2001At&T Corp.Low-latency buffering for packet telephony
US6356545 *Dec 12, 1997Mar 12, 2002Clarent CorporationInternet telephone system with dynamically varying codec
US6374225 *Oct 9, 1998Apr 16, 2002Enounce, IncorporatedMethod and apparatus to prepare listener-interest-filtered works
US6445696 *Feb 25, 2000Sep 3, 2002Network Equipment Technologies, Inc.Efficient variable rate coding of voice over asynchronous transfer mode
US6469737 *Jul 22, 1997Oct 22, 2002Canon Kabushiki KaishaImage-sensing server and control method and storage medium therefor
US6522730 *Jan 15, 1999Feb 18, 2003Texas Instruments IncorporatedDSL communication system with improved bandwidth
US6553568 *Sep 29, 1999Apr 22, 20033Com CorporationMethods and systems for service level agreement enforcement on a data-over cable system
US6556578 *Apr 14, 1999Apr 29, 2003Lucent Technologies Inc.Early fair drop buffer management method
US6580694 *Aug 16, 1999Jun 17, 2003Intel CorporationEstablishing optimal audio latency in streaming applications over a packet-based network
US6625655 *May 4, 1999Sep 23, 2003Enounce, IncorporatedMethod and apparatus for providing continuous playback or distribution of audio and audio-visual streamed multimedia reveived over networks having non-deterministic delays
US6625656 *Mar 6, 2000Sep 23, 2003Enounce, IncorporatedMethod and apparatus for continuous playback or distribution of information including audio-visual streamed multimedia
US6625711 *Nov 16, 2000Sep 23, 2003Cypress Semiconductor Corp.Method and/or architecture for implementing queue expansion in multiqueue devices
US6665728 *Dec 30, 1998Dec 16, 2003Intel CorporationEstablishing optimal latency in streaming data applications that use data packets
US6665751 *Apr 17, 1999Dec 16, 2003International Business Machines CorporationStreaming media player varying a play speed from an original to a maximum allowable slowdown proportionally in accordance with a buffer state
US6671292 *Jun 25, 1999Dec 30, 2003Telefonaktiebolaget Lm Ericsson (Publ)Method and system for adaptive voice buffering
US6683889 *Nov 15, 1999Jan 27, 2004Siemens Information & Communication Networks, Inc.Apparatus and method for adaptive jitter buffers
US6697356 *Mar 3, 2000Feb 24, 2004At&T Corp.Method and apparatus for time stretching to hide data packet pre-buffering delays
US6724776 *Nov 23, 1999Apr 20, 2004International Business Machines CorporationMethod and system for providing optimal discard fraction
US6804244 *Dec 14, 1999Oct 12, 2004Texas Instruments IncorporatedIntegrated circuits for packet communications
US20020080886 *Apr 4, 2001Jun 27, 2002Ptasinski Henry S.Method for selecting frame encoding parameters in a frame-based communications network
US20020089931 *Jul 11, 2001Jul 11, 2002Syuji TakadaFlow controlling apparatus and node apparatus
US20020105932 *May 24, 2001Aug 8, 2002Kazuyuki MiyaRadio communication system and communication terminal apparatus used therein
US20020114285 *Feb 26, 2002Aug 22, 2002Leblanc WilfridData rate controller
US20040045030 *Sep 26, 2002Mar 4, 2004Reynolds Jodie LynnSystem and method for communicating media signals
US20040062201 *Sep 30, 2002Apr 1, 2004Sharp Laboratories Of America, Inc.System and method for streaming TCP messages in an enterprise network
Non-Patent Citations
Reference
1A. Demers, et al., "Analysis and Simulation of a Fair Queueing Algorithm," Internetworking: Research and Experience, vol. 1, 1990, pp. 3-26.
2A. Rijsinghani, "Computation of the Internal Checksum via Incremental Update", IETF RFC 1624, May 1994.
3B. Braden, et al., "Recommendations on Queue Management and Congestion Avoidance in the Internet", IETF RFC 2309, Apr. 1998.
4L. Kalampoukas, et al., "Explicit Window Adaptation: A Method to Enhance TCP Performance", Proc. IEEE INFOCOM'97, San Francisco, CA, Mar. 1997, pp. 243-251.
5P. Narvaez, et al., "An Acknowledgment Bucket Scheme for Regulating TCP Flow Over ATM", Proc. Globecom '97, Phoenix, Arizona, 1997.
6S. Floyd, "TCP and Explicit Congestion Notification", Computer Commun. Revv. vol. 24, No. 5, Oct. 1994, pp. 8-23.
7S. Floyd, et al., "Random Early Detection Gateways for Congestion Avoidance", IEEE/ACM Trans, Networking, vol. 1, No. 4, Aug. 1993, pp. 397-413.
8Satyovolu, K. Duvedi, et al. "Explicit Rate Control of TCP Applications", ATM Forum Document, ATM-Forum/98-0152R1, Feb. 1998.
9T. Ziegler, et al., "Congestion Avoidance With BUC (Buffer Utilization Control) Gateways and RFCN (Reverse Feedback Congestion Notification)," IPCCC'97, Scottsdale, Arizona 1997.
10V. Jacobson, et al., "Congestion Avoidance and Control", Proc. ACM SIGCOMM'88, pp. 314-329.
11W. Stevens, "TCP Slow Start, Congestion Avoidance, Fast Retransmit, and Fast Recovery Algorithms", IETF RFC 1002, Jan. 1997.
Referenced by
Citing PatentFiling datePublication dateApplicantTitle
US7245586 *Aug 30, 2002Jul 17, 2007Lucent Technologies Inc.Buffer management based on buffer sharing across ports and per-port minimum buffer guarantee
US7349337 *Dec 12, 2003Mar 25, 2008Novell, Inc.Techniques for shaping data transmission rates
US7388834 *Jun 28, 2001Jun 17, 2008Hewlett-Packard Development Company, L.P.System and method for controlling network traffic flow in a multi-processor network
US7421502 *Dec 6, 2002Sep 2, 2008International Business Machines CorporationMethod and system for storage-aware flow resource management
US7484011 *Oct 8, 2003Jan 27, 2009Cisco Technology, Inc.Apparatus and method for rate limiting and filtering of HTTP(S) server connections in embedded systems
US7546367Jan 15, 2004Jun 9, 2009Novell, Inc.Methods and systems for managing network traffic by multiple constraints
US7583594 *Jan 31, 2003Sep 1, 2009Texas Instruments IncorporatedAdaptive transmit window control mechanism for packet transport in a universal port or multi-channel environment
US7627870 *Apr 28, 2001Dec 1, 2009Cisco Technology, Inc.Method and apparatus for a data structure comprising a hierarchy of queues or linked list data structures
US7672279 *Mar 6, 2003Mar 2, 2010Telefonaktiebolaget L M Ericsson (Publ)Methods for dynamic radio resource management and link control
US7706261 *Aug 27, 2004Apr 27, 2010Jinshen SunQueue-based active queue management process
US7706276 *Feb 28, 2007Apr 27, 2010Huawei Technologies Co., Ltd.Systems and methods for wireless communications
US7729249 *Nov 16, 2007Jun 1, 2010Microsoft CorporationSystems and methods for improving TCP-friendliness of delay-based congestion control
US7782885 *Dec 10, 2003Aug 24, 2010Cisco Technology, Inc.System and method for queue management using queue sets
US7788389 *Jun 6, 2008Aug 31, 2010International Business Machines CorporationProgram storage device for storage-aware flow resource
US7881725 *Jun 30, 2006Feb 1, 2011Nokia CorporationMethod and apparatus for providing adaptive thresholding for adjustment to loading conditions
US7912951 *Oct 28, 2008Mar 22, 2011Vmware, Inc.Quality of service management
US7948887 *Jun 24, 2008May 24, 2011Microsoft CorporationNetwork bandwidth measurement
US7983159Apr 12, 2010Jul 19, 2011Intellectual Ventures Holding 57 LlcQueue-based active queue management process
US8086731 *Jul 9, 2008Dec 27, 2011Hangzhou H3C Technologies Co., Ltd.Method, system and apparatus for collecting statistics of characteristic value with threshold
US8089878 *Jun 5, 2009Jan 3, 2012Fahd AlharbiAllocating bandwidth in a resilient packet ring network by P controller
US8125904 *May 30, 2006Feb 28, 2012Broadcom CorporationMethod and system for adaptive queue and buffer control based on monitoring and active congestion avoidance in a packet network switch
US8127014 *Jan 20, 2011Feb 28, 2012Vmware, Inc.Quality of service management
US8169909 *Feb 2, 2006May 1, 2012Nokia CorporationOptimization of a transfer layer protocol connection
US8200188Feb 20, 2009Jun 12, 2012Bridgewater Systems Corp.System and method for adaptive fair usage controls in wireless networks
US8230106 *Mar 31, 2003Jul 24, 2012Alcatel LucentMethods and apparatus for improved transmission control protocol transmission over a wireless channel exhibiting rate and delay variations
US8250197 *Oct 28, 2008Aug 21, 2012Vmware, Inc.Quality of service management
US8310930 *Jun 5, 2009Nov 13, 2012New Jersey Institute Of TechnologyAllocating bandwidth in a resilient packet ring network by PI controller
US8320246Feb 18, 2010Nov 27, 2012Bridgewater Systems Corp.Adaptive window size for network fair usage controls
US8374091 *Mar 26, 2009Feb 12, 2013Empire Technology Development LlcTCP extension and variants for handling heterogeneous applications
US8451852 *Sep 1, 2010May 28, 2013Juniper Networks, Inc.Systems and methods for selectively performing explicit congestion notification
US8520678 *Apr 18, 2011Aug 27, 2013Microsoft CorporationNetwork bandwidth measurement
US8527706 *Sep 16, 2011Sep 3, 2013Numecent Holdings, Inc.Opportunistic block transmission with time constraints
US8577329May 4, 2010Nov 5, 2013Bridgewater Systems Corp.System and methods for carrier-centric mobile device data communications cost monitoring and control
US8705367 *Jan 10, 2013Apr 22, 2014Empire Technology Development LlcTCP extension and variants for handling heterogeneous applications
US8717890 *Jan 29, 2010May 6, 2014Movik NetworksApplication, usage and radio link aware transport network scheduler
US8755405Nov 9, 2010Jun 17, 2014Movik Networks, Inc.Burst packet scheduler for improved ran efficiency in UMTS/HSPA networks
US8780705 *Apr 30, 2009Jul 15, 2014Freescale Semiconductor, Inc.Apparatus, communications system and method for optimizing data packet flow
US8838257 *Aug 1, 2008Sep 16, 2014Marvell International Ltd.Controller and design support apparatus
US20100106816 *Oct 28, 2008Apr 29, 2010Vmware, Inc.Quality of service management
US20100195602 *Jan 29, 2010Aug 5, 2010Movik NetworksApplication, Usage & Radio Link Aware Transport Network Scheduler
US20100246398 *Mar 26, 2009Sep 30, 2010Mung ChiangTcp extension and variants for handling heterogeneous applications
US20100290415 *May 6, 2010Nov 18, 2010Samsung Electronics Co., Ltd.Apparatus and method for bandwidth request in broadband wireless communication system
US20100322075 *Sep 1, 2010Dec 23, 2010Juniper Networks, Inc.Systems and methods for selectively performing explicit congestion notification
US20110194446 *Apr 18, 2011Aug 11, 2011Microsoft CorporationNetwork bandwidth measurement
US20120039174 *Apr 30, 2009Feb 16, 2012Freescale Semiconductor, Inc.Apparatus, communications system and method for optimizing data packet flow
US20120278555 *Sep 16, 2011Nov 1, 2012Endeavors Technologies, Inc.Opportunistic block transmission with time constraints
US20130121148 *Jan 10, 2013May 16, 2013Empire Technology Development LlcTcp extension and variants for handling heterogeneous applications
CN1976343BNov 10, 2006Jul 28, 2010华为技术有限公司Method and system for raising transmission control protocol data handling capacity
EP2247044A1 *Apr 29, 2009Nov 3, 2010Alcatel LucentPacket network processing node and method
WO2010095036A1 *Feb 19, 2010Aug 26, 2010Bridgewater Systems Corp.Adaptive window size for network fair usage controls
WO2010095038A1 *Feb 19, 2010Aug 26, 2010Bridgewater Systems Corp.System and method for adaptive fair usage controls in wireless networks
WO2012163879A1 *May 29, 2012Dec 6, 2012Alcatel LucentMethod and apparatus for mitigating buffering delay at a network edge
Classifications
U.S. Classification709/235, 370/231
International ClassificationG06F15/16
Cooperative ClassificationH04L69/163, H04L69/16, H04L47/29, H04L47/35, H04L47/2441, H04L47/27, H04L47/25, H04L1/1832, H04L47/193, H04L47/30, H04L47/10
European ClassificationH04L1/18R1, H04L29/06J7, H04L47/30, H04L47/24D, H04L47/25, H04L47/10, H04L47/19A, H04L47/27, H04L47/29, H04L47/35, H04L29/06J
Legal Events
DateCodeEventDescription
Oct 16, 2013FPAYFee payment
Year of fee payment: 8
Feb 26, 2010ASAssignment
Owner name: AVAYA HOLDINGS LIMITED,NEW JERSEY
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:NORTEL NETWORKS LIMITED;US-ASSIGNMENT DATABASE UPDATED:20100312;REEL/FRAME:23998/799
Effective date: 20091218
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:NORTEL NETWORKS LIMITED;US-ASSIGNMENT DATABASE UPDATED:20100301;REEL/FRAME:23998/799
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:NORTEL NETWORKS LIMITED;US-ASSIGNMENT DATABASE UPDATED:20100316;REEL/FRAME:23998/799
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:NORTEL NETWORKS LIMITED;US-ASSIGNMENT DATABASE UPDATED:20100413;REEL/FRAME:23998/799
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:NORTEL NETWORKS LIMITED;REEL/FRAME:023998/0799
Owner name: AVAYA HOLDINGS LIMITED, NEW JERSEY
Oct 23, 2009FPAYFee payment
Year of fee payment: 4
Dec 18, 2000ASAssignment
Owner name: NORTEL NETWORKS LIMITED, CANADA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:AWEYA, JAMES;OUELLETTE, MICHEL;MONTUNO, DELFIN Y.;REEL/FRAME:011384/0769;SIGNING DATES FROM 20001208 TO 20001213